Both the Swedish PSA database T-Book and the German PSA database ZEDB use two-stage Bayesian models to calculate plant specific and generic reliability data from operational experience (component event data and operating times) of nuclear power plants.
Generally speaking, different assumptions can be made for the hyper distribution and hyper priors for the parameters of the hyper distribution, and these may influence the calculated reliability data. Evaluating the same data sets using the T-Book and the ZEDB approaches provides the opportunity to compare the methodologies.
Therefore TUD (responsible for the T-Book) and VGB (responsible for the ZEDB) agreed on a benchmark, the subject of which was to apply the T-Book/ZEDB statistical tools to real data to observe the consequences for the estimated failure rates. For this purpose T-Book-data sets are evaluated using the ZEDB approach and ZEDB-data sets using the T-Book approach.
Pörn Consulting performed the evaluation of the ZEDB event-data on behalf of TUD and Framatome ANP evaluated the T-Book-data sets on behalf of VGB.
The publication Calculation of Reliability Data using Two-Stage Bayesian Models: T-Book ZEDB Benchmark (TW 820e) presents the results of the benchmark; Appendix 1 contains the T-Book evaluation, Appendix 2 the ZEDB evaluation, Appendix 3 a comparison of T-Book calculation of ZEDB data sets with original ZEDB results, and Appendix 4 the exchanged T-Book/ZEDB data sets.
This publication will be delivered as an ebook (PDF-file).
